TadPose

Automated behavioural phenotyping of Xenopus laevis tadpoles from 24-well plate video.

TadPose turns hour-long multi-well recordings of tadpoles into a data-driven catalogue of behavioural prototypes, enabling unsupervised quantification of seizure phenotypes in models of developmental and epileptic encephalopathies (DEE). It combines multi-landmark posture dynamics with centre-of-mass kinematics, clusters them without human-defined categories, and compares the resulting behavioural repertoire across drug and genetic seizure models.

The pipeline uses DeepLabCut for markerless pose estimation and reRandomStats for re-randomisation statistics.

Contents

• Overview
  • What problem does TadPose address?
  • The pipeline
  • What the method recovers
  • Provenance
• Installation
  • Environment
  • Dependencies
  • Configure machine-specific paths (required first step)
• Usage
  • Command-line interface
  • Choosing the number of clusters
  • Using the library
  • High-performance computing
• API reference
  • Configuration
  • Acquisition and pose estimation
  • Feature extraction
  • Clustering
  • Cluster validation (choosing k)
  • Post-clustering analysis
  • Visualisation
  • Database and experiment management

Indices

• Index

• Module Index

• Search Page